Viability, Electability, and Candidate Choice in a Presidential Primary Election: A Test of
Competing Models
Author(s): Alan I. Abramowitz
Source: The Journal of Politics, Vol. 51, No. 4, (Nov., 1989), pp. 977-992
Published by: Cambridge University Press on behalf of the Southern Political Science
Association
Stable URL: http://www.jstor.org/stable/2131544
Accessed: 22/05/2008 07:33
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=cup.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We enable the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].
http://www.jstor.org
Viability,Electability,and Candidate
Choicein a Presidential
PrimaryElection:A Test of Competing
Models
Alan I. Abramowitz
Emory University
Using data from an exit poll, this paper tests three models of voter decision making in a presidential primary: a simple candidate preference model, a bandwagon model, and an expected
utility model. For both Republican and white Democratic primary voters, the data support the
expected utility model. In choosing a candidate for their party's nomination, Republican and
Democratic primary voters weighed electability in addition to their general evaluations of the
candidates. Opinions about electability were, in turn, strongly influenced by perceptions of candidates' nomination prospects. Thus, viability had an important, but indirect, influence on voter
decision making.
presidential primary election presents voters with a decision-making
context which differs sharply from that of a general election or even another
type of primary election. Perhaps the most important difference between
presidential primaries and other types of primary elections is the fact that
presidential primaries involve a series of separate contests spread over several months. Thus, the results of early contests, and the way these results are
interpreted by the media, can influence voters in later contests. The New
Hampshire primary and, in recent years, the Iowa precinct caucuses, have
been considered especially significant in this regard because they are the
first major events of the nominating campaign, and they receive extremely
heavy coverage from the news media (Orren and Polsby 1987).
The sequential character of the presidential nominating process and the
intense coverage which the news media devote to Iowa and New Hampshire
have led to a great deal of speculation about the significance of momentum in
presidential nominating campaigns. By winning or doing "better than expected" in Iowa and New Hampshire, a presidential candidate is said to gain
momentum in subsequent contests. Thus, in 1976, Jimmy Carter, a little
known ex-Governor of Georgia, finished first in Iowa and New Hampshire
and went on to win the Democratic nomination. Eight years later, Gary Hart
emerged from obscurity to challenge Walter Mondale for the Democratic
A
Vol. 51, No. 4, November 1989
C)1989 by the University of Texas Press
JOURNAL OF POLITICS,
978
Alan I. Abramowitz
nomination on the basis of a "surprisinglystrong"second place finish in Iowa
and a victory in New Hampshire.
Momentum is widely regarded as a major factor in the presidential nominating process. However, very little is known about why momentum is important or how it affects voter decision making in primary elections. In fact,
there has been almost no systematic research on voter decision making in
primaries. A few studies have analyzed citizens' pre-nomination candidate
preferences. Bartels (1985), using data from the preconvention waves of the
1980 National Election Study, found a reciprocal relationship between citizens' expectations about the outcome of the nominating process and their
candidate preference. The effect of expectations on preferences appeared to
be strongest during the early stages of the nominating campaign.
Several recent studies have used data from the NES 1984 "rolling crosssection" survey to analyze pre-nomination candidate preferences. Abramowitz (1987), Bartels (1987), and Brady and Johnston (1987) all found that
opinions about Gary Hart's and Walter Mondale's chances of winning the
Democratic nomination had a significant influence on citizens' candidate
preferences. Evaluations of Hart's nomination prospects and support for his
candidacy both increased dramaticallyafter his victory over Mondale in New
Hampshire.
These findings appear to be consistent with the momentum hypothesis.
However, the findings from the "rolling cross-section" survey may have reflected circumstances peculiar to the 1984 Democratic campaign-the fact
that Gary Hart was largely unknown before the Iowa and New Hampshire
contests may have magnified the impact of his early successes in those states.
Moreover, none of these studies involved actual primary voters. Voters in a
state holding a primary election may have a fuller opportunity to evalute the
entire field of candidates than citizens in states where the candidates have
not been campaigning. Direct exposure to the candidates and their campaigns may reduce the impact of momentum on primary voters.
THREE MODELS OF VOTER DECISION MAKING IN PRESIDENTIAL
PRIMARIES
This paper will consider three models of voter decision making in presidential primaries. In each model, the dependent variable is the voter's candidate preference (Choice); the independent variables are the voter's overall
evaluations of the major candidates (Candidate Evaluation), the voter's perceptions of the candidates' chances of receiving their party's nomination
(Viability), and the voter's perceptions of the candidates' chances of winning
the November election (Electability).
Although many other variables, including social background characteristics and policy preferences, may have affected voters' decisions in the primary, previous research on voting behavior in general elections has shown
979
Viability, Electability, and Candidate Choice
FIGURE 1
*THREE MODELS OF VOTER DECISION
MAKING IN PRESIDENTIAL PRIMARIES
A. Simple Candidate Choice Model
Viability
Candidate
Choice
Evaluation
Electability
B. Bandwagon Model
Viability
Candidate
Evaluation
Io
Choie
Electability
C. Expected Utility Model
Viability
Candidate
Evaluation
I
Choice
Choice
Electability
that the effects of these additional variables are almost entirely indirectmediated by evaluations of the candidates (Markus and Converse 1979;
Markus 1982). There is no reason to expect candidate evaluations to play a
less central role in voter decision making in primary elections. In fact, given
the absence of partisan cues, candidates are probably even more salient in
primaries than in general elections. As long as the effects of social background characteristics and policy preferences are mediated by candidate
evaluations, then leaving these variables out of the analysis should not bias
our estimates of the effects of the variables included in the model.
All three models assume that voters engage in wishful thinking-opinions
about the candidates' chances of winning the nomination and the general
election are based in part on voters' evaluations of the candidates. However,
opinions about the candidates' nomination prospects should also reflect the
980
Alan I. Abramowitz
results of earlier primaries and the media's interpretations of these results.
All three models also assume that judgments about electability are influenced by voters' opinions about the candidates' nomination prospects. The
most direct evidence that voters have available about a candidate's ability to
wage an effective general election campaign is his or her ability to wage an
effective pre-nomination campaign. Therefore, a candidate who does well or
better than expected in the early primaries and caucuses will probably be
viewed as more electable than a candidate who does poorly or worse than
expected (see Aldrich 1980, 80-82).
The first model of voter decision making which will be considered in this
paper is a simple candidate preference model. According to this model,
opinions about the candidates' nomination chances and electability have no
effects on voters' candidate preferences (see figure 1). Voters choose the candidate they evaluate most positively, and they also tend to assume that the
candidate they like best is the one most likely to win the nomination and the
general election. In this model, the results of earlier primaries and caucuses
and media coverage of these results are important only if they influence
voters' evaluations of the candidates.
The second model which will be considered in this paper is a bandwagon
model. According to this model, opinions about the candidates' nomination
chances directly influence voters' candidate preferences, but opinions regarding electability have no effect on candidate preferences. The motivational assumption underlying this model is that voters want to be on the
winning side in the nominating campaign because supporting a winner is intrinsically more enjoyable than supporting a loser. However, voters in this
model are concerned exclusively about the nominating stage of the presidential selection process-they do not weigh electability as a separate criterion in choosing a candidate.
The third model which will be considered in this paper is an expected utility model. According to this model, primary voters weigh electability along
with their evaluations of the candidates in making a choice. The assumption
underlying this model is that primary voters are rational actors who seek to
maximize their expected utility (Aldrich 1980, 80-82). Candidate evaluations, in this model, represent voters' assessments of the utility which they
would obtain if the candidates were elected to the presidency. Therefore,
these evaluations must be discounted by the subjective probability of each
candidate winning the general election. The results of earlier primaries and
caucuses are important primarily because they provide evidence about the
candidates' chances of winning the general election.
Data and Methodology
The data used to test these three models of voter decision making come
from an exit poll of presidential primary voters conducted in Dekalb County,
Viability, Electability, and Candidate Choice
981
Georgia, on March 8, 1988. Self-administered questionnaries were completed by 451 Democratic and Republican primary voters at 13 randomly
selected precincts. Of the 451 respondents in the exit poll, 278 reported
voting in the Democratic primary and 173 reported voting in the Republican
primary. Despite the relatively small size of this sample of primary voters,
the candidate preferences of the respondents in the exit poll came very close
to matching the preferences of all Democratic and Republican primary
voters in Dekalb County. More detailed information about the procedures
used to conduct the exit poll and the validity of the sample are provided in
appendix A.
For the purpose of analyzing voting behavior in a primary election, an exit
poll has several major advantages over conventional survey techniques. It is
possible to measure candidate preferences and other attitudes immediately
after voters have cast their ballots, before these attitudes are contaminated
by information about the results of the primary. Conventional survey techniques (either telephone or personal interviews) require that citizens be interviewed either before the primary, when they may not have reached a final
decision, or after the primary, when their attitudes may have been modified
by exposure to information about the results of the primary. This problem is
especially serious when it comes to measuring voters' opinions about candidates' nomination prospects and electability. In addition, all of the respondents in an exit poll are actual primary voters. With conventional survey
techniques, it is necessary to determine which respondents are likely to vote
or, in a post-election survey, which respondents actually did vote. Since the
level of voter turnout in presidential primary elections is usually quite low
(averaging less than one-third of the voting-age population in recent years),
the task of identifying actual voters in a cross-sectional survey is quite
problematic.
The principal drawback of exit poll data is that the number of questions
which can be asked is very limited, and the questions must be kept very
simple so that voters can complete the questionnaire in a few minutes. In
addition to asking respondents which candidate they voted for in the primary, the Dekalb County Exit Poll included questions asking voters for an
overall evaluation of each majorcandidate in both parties, and for their opinions about which candidates had the best chance of winning the Democratic
and Republican nominations, and which candidate in each party had the best
chance of winning the November election if nominated by his party. These
questions were used to analyze the effects of candidate evaluations, momentum, and electability on voting decisions in the primary.1
Path regression analysis was used to test the three models of momentum
in a presidential primary. The advantage of path analysis is that the direct
'See appendix A for the wording of each of these questions, and the coding procedures used
in the regression analyses.
982
Alan I. Abramowitz
and indirect effects of candidate evaluations, viability, and electability on
candidate choice can be estimated (Asher 1976). Although the use of a dichotomous dependent variable violates some of the assumptions of regression analysis, the consequences of these violations are generally not severe
unless the dependent variable has a very skewed distribution (Aldrich and
Cnudde 1975). Since this was not the case, ordinary regression analysis was
used to estimate the effects of our independent variables. Discriminant analyses of candidate preference were conducted for both Democratic and Republican primary voters, and the results were very similar to those of the
regression analyses. These results are summarized in appendix B.
The Setting: Super Tuesday
On March 8, 1988, presidential primaries were held in 15 southern and
border states, including Georgia, along with several states outside of the
South. The brainchild of a group of moderate southern Democratic party
leaders, "Super Tuesday" was designed to attract more attention to the
South from the presidential candidates and to improve the prospects of a
moderate candidate in the Democratic party. The Super Tuesday primaries
were the first major campaign event following the New Hampshire primary
which was held on February 16. Thus, it should be possible to analyze the
impact of the New Hampshire results on Super Tuesday primary voters.
The New Hampshire primary reestablished George Bush as the frontrunner for the Republican presidential nomination following a poor third-place
finish in the Iowa precinct caucuses. Pre-Iowa polls had shown Bush with a
huge lead over Robert Dole and Pat Robertson in New Hampshire. Following Dole's victory in Iowa, however, the polls and pundits generally portrayed the New Hampshire race as a toss-up between Bush and Dole. Although Bush's eight percentage point margin over Dole was actually much
smaller than his earlier lead in the polls, the New Hampshire result was
widely interpreted by the news media as a major disappointment and setback to Dole's campaign. In fact, Dole's campaign never recovered from his
loss in New Hampshire. On Saturday, March 5, just three days before Super
Tuesday, George Bush won a decisive victory over Robert Dole and Pat
Robertson in the South Carolina Republican primary. On Super Tuesday,
Bush won every Republican primary and virtually clinched the GOP
nomination.
Michael Dukakis had been viewed as the early favorite in the New Hampshire Democratic primary because of his position as the governor of neighboring Massachusetts. However, after Dukakis finished third in Iowa behind
RichardGephardt and Paul Simon, there was some question about his ability
to maintain his lead in New Hampshire. Dukakis' decisive victory in New
Hampshire, combined with a huge financial and organizational advantage
over the other Democratic candidates, established him as the Democratic
Viability, Electability, and Candidate Choice
983
frontrunner going into Super Tuesday. Dukakis' position as the Democratic
frontrunner was rather tenuous however. His ability to appeal to voters in
the South was uncertain, and he faced two additional candidates on Super
Tuesday who had not been major factors in Iowa and New Hampshire: Jesse
Jackson and Albert Gore, Jr. The outcome of Super Tuesday on the Democratic side was a three-way split in both the popular vote and the delegate
race among Dukakis, Jackson, and Gore. However, by finishing first in Texas
and Florida and by maintaining his overall lead in delegates, Michael Dukakis probably solidified his position as the frontrunner for the Democratic
nomination.
Results
There was a three-way split among Democratic primary voters in the exit
poll, with Jesse Jackson winning 40% of the vote, followed by Albert Gore,
Jr. with 26% and Michael Dukakis with 24%. The remaining 10% of the
Democratic primary vote was split among Paul Simon (4%), Richard Gephardt (3%), Gary Hart (2%), and uncommitted delegates (1%). Among Republican primary voters in the exit poll, George Bush received 51% of
the vote compared with 29% for Robert Dole, 11% for Pat Robertson, and
7% for Jack Kemp. Pierre DuPont and Alexander Haig each received 1% of
the vote.
There was a high level of agreement among Democratic and Republican
primary voters about which candidate had the best chance of winning each
party's nomination. Michael Dukakis was perceived as the frontrunner for
the Democratic nomination by 58% of Democratic primary voters. Trailing
Dukakis among Democratic primary voters were Jesse Jackson at 21%, Albert Gore at 14%, and Richard Gephardt at 6%. Dukakis was also seen as the
Democratic frontrunner by 64% of Republican primary voters, followed by
Gephardt and Gore with 15% each, and Jacksonwith 4%.
George Bush enjoyed an even greater advantage in the GOP contest: 77%
of Democratic primary voters and 80% of Republican primary voters viewed
Bush as the most likely Republican nominee. Twenty percent of Democratic
primary voters and 17%of Republican primary voters picked Robert Dole as
the most likely GOP nominee. Only 2% of Democratic primary voters and
4% of Republican primary voters thought that Pat Robertson had the best
chance of winning the Republican presidential nomination.
The question of electability also produced a strong consensus across party
lines. Dukakis was viewed as the most electable Democratic candidate by
56% of Democratic primary voters and 63% of Republican primary voters.
Among Democratic voters, Dukakis was followed by Jacksonat 21%, Gore at
16%, and Gephardt at 5%;among Republican voters, 17%picked both Gore
and Gephardt as the most electable Democrat while only 2% chose Jesse
Jackson.
984
Alan I. Abramowitz
George Bush was viewed as the most electable Republican candidate by
77% of Democratic primary voters and by 74% of Republican primary
voters. Among Democratic voters, Robert Dole was viewed as the most electable Republican candidate by 22% while Jack Kemp and Pat Robertson
were each picked by 1%; among Republican voters, Dole trailed Bush at
23%, followed by Robertson at 3% and Kemp at 1%.
Judgments about candidates' nomination and general election prospects
were clearly more than rationalizationsof candidate preferences: Democrats
and Republicans alike tended to view Michael Dukakis and George Bush as
the frontrunners for the Democratic and Republican nominations and as the
strongest potential candidates in the general election. This was true despite
the fact that, among Democratic primary voters, Robert Dole was evaluated
much more positively than George Bush-42% of Democratic primary
voters had a favorable opinion of Dole while only 24% had a favorable opinion of Bush. The fact that far more voters rated Dukakis and Bush as having
the best chance of being nominated and elected than actually voted for
either candidate suggests that opinions about the candidates' nomination
prospects and electability were strongly influenced by the results of earlier
contests and by media interpretations of these results.
How did the widespread perception of Michael Dukakis as the frontrunner for the Democratic nomination affect the candidate preferences of
Democratic primary voters on Super Tuesday? In order to answer this question, it is first necessary to control for the influence of race. According to our
exit poll, blacks comprised 35% of the Democratic primary electorate and
Jesse Jackson received 97% of the black vote compared with 2% for Albert
Gore and 1% for Michael Dukakis. Among whites who voted in the Democratic primary, Gore received 39% of the vote followed by Dukakis with
38%, Jackson and Simon with 7% each, Hart with 3%, and Bruce Babbitt
and uncommitted delegates with 1% each.
Because of the overwhelming support for Jesse Jackson among black
voters, our analysis of the effect of momentum on candidate choice will be
limited to white voters in the Democratic primary. Among blacks, support
for Jesse Jacksonwas a matter of racialpride.2 We will further limit our attention to the two candidates who received the overwhelming majority of the
white vote-Michael Dukakis and Albert Gore.
Figure 2 presents the results of the path analysis of candidate preference
among white Democratic primary voters. The estimates shown are the standardized regression coefficients or beta weights. These results support the
expected utility model of voter decision making. Although candidate evaluations had the strongest direct influence on voting decisions, judgments about
2A separate analysis of the Jacksonvote found that race was, by far, the strongest predictor of
voting for Jacksonfollowed by evaluations of Jackson. Opinions about Jackson'snomination prospects and electability had no impact on voting decisions.
Viability, Electability, and Candidate Choice
FIGURE
985
2
PATH ANALYSIS OF CANDIDATE CHOICE IN DEMOCRATIC PRIMARY
98)
(N=
Viability
.389-7
Dukakis
Evaluation
.040*
-.332
.446
-.464
-.020
Gore
Evaluation
.
Choice
.580
Electability
Estimated Effects of Independent Variables
Direct
Dukakis evaluation
.446
-.464
Gore evaluation
.040
Viability
.237
Electability
Indirect
.135
-.100
.137
Note: Estimates shown are standardized regression coefficients. Double-headed arrow represents correlation coefficient. Estimate marked with asterisk is not statistically significant. All
other estimates are significant at .05 level or greater.
which candidate had the best chance of winning the November election also
had a substantial direct impact on candidate choice. Based on the estimated
unstandardized regression coefficient for electability, perceiving Michael
Dukakis as the most electable Democratic candidate increased the probability of voting for Dukakis over Albert Gore by about 27 percentage points
compared with perceiving Gore as the most electable Democratic candidate.
By multiplying the estimated unstandardized regression coefficient for relative electability (.135) by the mean electability score among white Democratic primary voters (+ .41), we can estimate that perceptions of electability
increased Dukakis' overall vote by about 5.5 percentage points over what he
would have received if he and Gore had been perceived as equally electable.
Opinions about the candidates' viability apparently had little or no direct
influence on voting decisions, but did strongly influence opinions about
electability. The estimated unstandardized regression coefficient for the
effect of viability on electability (.601) indicates that perceiving either
Michael Dukakis or Albert Gore as the most likely Democratic nominee increased the likelihood of perceiving the same candidate as the most electable
Democrat by about 60 percentage points.
There is clear evidence of wishful thinking among Democratic primary
voters: voters tended to perceive the candidate they liked best as having the
best chance of being nominated and elected. However, perceptions of the
Alan I. Abramowitz
986
FIGURE
PATH ANALYSIS
OF CANDIDATE
3
CHOICE
IN REPUBLICAN
PRIMARY
(N = 128)
Viability
Bush
Evaluation
-.081*
.246
.406
-.342
-.297
Dole
Evaluation
Choice
.744
-
\
2 r
.4~215
/
Electability
Estimated Effects of Independent Variables
Direct
Bush evaluation
.406
Dole evaluation
-.297
Viability
-.081
Electability
.415
Indirect
.154
-.110
.309
Note: Estimates shown are standardized regression coefficients. Double-headed arrow represents correlation coefficient. Estimate marked with asterisk is not significant at .05 level. All
other estimates are statistically significant at .05 level or greater.
candidates' nomination prospects were the most important factor influencing
opinions about electability, and evaluations of Dukakis and Gore only explained about one-fourth of the variance in voters' opinions about which candidate was most likely to receive the Democratic nomination.
Among Republican primary voters in our exit poll, George Bush and
Robert Dole received a combined total of 80% of the vote. We will therefore
limit our analysis of Republican primary voters to those choosing one of
these two candidates. Figure 3 presents the results of the path analysis of
candidate preference among Republican primary voters. Once again, the results are consistent with the expected utility model of voter decision making.
Among Republican primary voters, moreover, opinions.about the candidates' electability had a stronger direct impact on voting decisions than
evaluations of either of the candidates. Based on the estimated unstandardized regression coefficient for electability, perceiving George Bush as the
most electable Republican candidate increased the probability of voting for
Bush over Robert Dole by about 48 percentage points compared with perceiving Dole as the most electable Republican candidate. By multiplying the
estimated unstandardized regression coefficient for relative electability
(.238) by the mean electability score among Republican primary voters
(+.55), we can estimate that perceptions of electability increased Bush's
Viability, Electability, and Candidate Choice
987
overall vote by about 13.1 percentage points over what he would have received if he and Dole had been perceived as equally electable.
Perceptions of the candidates' nomination prospects had no direct impact
on voting decisions (in fact, the estimate for this path is in the wrong direction), but did strongly influence opinions about electability. The estimated
unstandardized regression coefficient for the effect of viability on electability
(.812) indicates that perceiving either George Bush or Robert Dole as the
most likely Republican nominee increased the likelihood of perceiving the
same candidate as the most electable Republican by about 81 percentage
points. Wishful thinking is again evident although evaluations of Bush and
Dole had very little direct bearing on opinions regarding electability and
only explained about one-third of the variance in judgments about which
candidate was most likely to win the Republican nomination.
An alternative explanation for these findings is that they reflect post-decisional rationalization-after voters decide which candidate to support, they
may rationalize that decision by assuming that their preferred candidate is
the one most likely to win the nomination or the general election or both.
However, the evidence from our exit poll appears to be inconsistent with
this hypothesis. If voters rationalized their candidate preference by assuming that their preferred candidate was most likely to win the nomination and
the general election, then candidate preference should have a direct influence on perceptions of the candidates' nomination prospects, after controlling for perceptions of electability. However, when perceptions of nomination prospects were regressed on candidate choice and perceptions of
electability, vote choice had almost no impact on opinions regarding the candidates' nomination prospects. If voters' candidate preferences only affected
their judgments about electability, then opinions about the candidates' nomination prospects should have no impact on voting decisions with electability
left out of the analysis. However, opinions about the candidates' nomination
prospects did have substantial and statistically significant effects on candidate preference in both parties when electability was excluded from the regression equation. Thus, if we can assume that opinions about candidates'
nomination prospects are causally prior to opinions about their electability,
the data from the exit poll are inconsistent with the post-decisional rationalization hyothesis.
DISCUSSION AND CONCLUSIONS
Momentum had a major influence on the voting decisions of Republican
and white Democratic primary voters in our exit poll. By winning the New
Hampshire primary, George Bush and Michael Dukakis established themselves as the frontrunners for the Republican and Democratic presidential
nominations in the minds of Super Tuesday primary voters. This perception
was important primarily because it caused voters to view Bush and Dukakis
988
Alan I. Abramowitz
as the most electable candidates in their respective parties. In choosing a
candidate for their party'snomination, Republican and Democratic primary
voters weighed electability in addition to their general evaluations of the
candidates seeking the nomination.
Although momentum was an important factor in both primaries, it probably was not the determining factor in the outcome of either. In the Republican primary, momentum reinforced the advantage in voter evaluations
which George Bush enjoyed over Robert Dole. Without momentum, however, Bush's margin over Dole would probably have been substantially
smaller. In the Democratic primary, the candidate with momentum,
Michael Dukakis, finished third behind Jesse Jacksonand Albert Gore. Dukakis' momentum was not enough to overcome Jackson'sstrong appeal to
black voters or Gore's emphasis on ideological moderation and his appeal
to regional loyalty. Without momentum, however, it is likely that Dukakis
would have fared very poorly on Super Tuesday.
The findings presented in this paper do not support the idea that momentum is especially important when the candidates seeking the nomination are
not well known. In our exit poll, the impact of momentum was greater in the
Republican primary than in the Democratic primary, despite the fact that
the two major Republican candidates-Bush and Dole-were very familiar
national political figures, while both Michael Dukakis and Albert Gore were
almost unknown at the start of the campaign. The importance of momentum
in the Republican race may have reflected the absence of any clear issue or
ideological differences between the two leading contenders. Both Bush and
Dole campaigned in the South as conservatives who strongly supported
Ronald Reagan'spolicies. The only controversy in the campaign was over
which candidate had been more effective in supporting Reagan. Lacking any
other basis on which to distinguish between Bush and Dole, Republican primary voters relied heavily on their judgment about which candidate had the
best chance to win in November. That judgment was, in turn, strongly influenced by the perception of George Bush as the clear frontrunner for the
GOP nomination.
To a considerable extent, the voters in our exit poll acted as rational utility
maximizers. Anticipating the upcoming general election, they weighed electability along with their evaluations of the candidates in deciding whom to
support in the primary. This finding may provide some comfort to those concerned about the effects of recent nominating reforms on the ability of the
parties to choose candidates with broad electoral appeal (Polsby 1983). However, students of the nominating process may find it somewhat disturbing
that primary voters base their evaluation of a candidate's electability almost
exclusively on his performance in earlier primaries and caucuses. On Super
Tuesday, George Bush and Michael Dukakis were seen as the most electable
Republican and Democratic candidates largely because of their victories in
the New Hampshire primary three weeks earlier. Many political analysts
Viability, Electability, and Candidate Choice
989
and professional politicians did not share this view, however. Several national polls conducted before Super Tuesday indicated that Robert Dole
would be a stronger Republican nominee than George Bush because of his
greater appeal to independent voters and Democrats. At the same time,
many southern Democratic party leaders and elected officials felt that Albert
Gore would be a stronger candidate than Michael Dukakis because of his
southern roots and moderate image. Judging a candidate's ability to win a
general election on the basis of his success in party primaries and caucuses
can be very misleading. Primaries and caucuses only test a candidate's appeal among voters who identify with the candidate'sparty. In order to win a
presidential election, however, a Democratic or Republican candidate must
also appeal to independents and supporters of the opposing party. The candidate with momentum in March is not necessarily the candidate with the best
chance of winning in November.
Much more research is needed on the role of momentum and electability
in presidential primary elections. It would be hazardousto attempt to generalize from the findings of an exit poll conducted in a single county in one
early primary contest. If possible, future studies should employ more sensitive measures of attitudes about candidates' nomination and general election prospects instead of asking voters which candidate is most likely to win.
By obtaining probability or quasi-probability estimates of candidates'
chances, it should be possible to test more sophisticated models of voter decision making (Abramowitzand Stone 1984, chap. 6). Future studies should
also explore the effects of campaign context and timing on voter decision
making. The relationship between momentum and voter perceptions of
viability and electability may change over time: the results of recent primaries may have less impact on perceptions of viability and electability in the
later stages of the nominating campaign. Finally, research is needed on the
role of media coverage in shaping voters' evaluations of candidate performance in the primaries and caucuses. Given the role that perceptions of momentum play in voter decision making, the power of the media to set expectations regarding a candidate's performance and to evaluate performance in
relation to these expectations may be crucial in determining a candidate's
success in the nominating campaign.
Manuscript submitted 10 August 1988
Final version received 3 April 1989
APPENDIX A
DESCRIPTION OF SAMPLING PROCEDURE, QUESTION WORDING, AND
CODING PROCEDURES USED IN EXIT POLL
A stratified random sample of 13 precincts in Dekalb County was used to
conduct the exit poll. Because of the high level of racial polarization ex-
990
Alan I. Abramowitz
pected in the Democratic primary, all of the precincts in the county were
first stratified according to racial composition. Within each racial grouping, a
random sample of precincts was selected. Questionnaires were distributed
and collected by undergraduate students at Emory University. Each student
was assigned a specific precinct and two-hour time period and instructed to
collect as many questionnaires as possible from voters leaving the polling
place. Time periods were distributed throughout the day, but concentrated
mainly during periods of heavy voting (early morning, midday, late afternoon, and early evening).
The results of the exit poll were very close to the actual results of the primary election in Dekalb County. In the county, 61% of the voters chose to
participate in the Democratic primary while 39% chose to participate in the
Republican primary;in the exit poll, 62% of the respondents reported voting
in the Democratic primary while 38% reported voting in the Republican primary. In the Democratic primary, Jesse Jackson received 46% of the vote
compared with 24% for Albert Gore, 23% for Michael Dukakis, 4% for
Richard Gephardt, 3% for Paul Simon, and 1% each for Gary Hart and uncommitted delegates; in the exit poll, 40% of respondents reported voting
for Jesse Jacksoncompared with 26% for Albert Gore, 24% for Michael Dukakis, 4% for Paul Simon, 3% for Richard Gephardt, 2% for Gary Hart, and
1%for uncommitted delegates. In the Republican primary, George Bush received 54% of the vote compared with 27% for Robert Dole, 12% for Pat
Robertson, and 7% for Jack Kemp; in the exit poll, 51% of respondents reported voting for George Bush with 29% for Robert Dole, 11% for Pat
Robertson, and 7% for Jack Kemp.
In the exit poll, candidate preference was measured by the following question: "Which candidate did you vote for in the primary?"Respondents were
given a checklist of all Democratic and Republican candidates.
Opinions about the candidates' nomination prospects were measured by
the following question: "Regardlessof whom you support, which Democratic
and Republican candidate do you think has the best chance of winning his
party'snomination?"Respondents were given checklists of the following candidates: Dukakis, Gephardt, Gore, Hart, Jackson, and Simon for the Democrats; Bush, Dole, Kemp, and Robertson for the Republicans.
Opinions about the candidates' general election chances were measured
by the following question: "Regardless of whom you support, which Democratic and Republican candidate do you think has the best chance of winning
the general election in November if he is nominated by his party?" The
choices were Dukakis, Gephardt, Gore, Hart, Jackson, and Simon for the
Democrats, and Bush, Dole, Kemp, and Robertson for the Republicans.
Overall evaluations of the candidates were measured by the following
question: "What is your overall opinion of each of the following political leaders?" Respondents were asked for their opinions of Bush, Dole, Dukakis,
991
Viability, Electability, and Candidate Choice
Gephardt, Gore, Jackson, and Robertson. The response alternatives were
"very favorable,""somewhatfavorable,""neutral,""somewhat unfavorable,""
and "very unfavorable."
In the path analysis of Democratic primary voters, the vote choice question was scored as + 1 for a Dukakis vote and 0 for a Gore vote. In the analysis of Republican primary voters, a Bush vote was scored as + 1 and a Dole
vote was scored as 0.
Candidate evaluations were scored from -2 (very unfavorable) to +2
(very favorable), with a neutral opinion scored as a 0. Among Democratic
primary voters, opinions about the candidates' nomination prospects and
electability were scored as follows: + 1 if Dukakis was viewed as having the
best chance, -1 if Gore was viewed as having the best chance, and 0 if some
other candidate was viewed as having the best chance.
Among Republican primary voters, opinions about the candidates' nomination prospects and electability were scored as follows: +1 if Bush was
viewed as having the best chance, -1 if Dole was viewed as having the best
chance, and 0 if some other candidate was viewed as having the best chance.
APPENDIX
B
RESULTS OF DISCRIMINANT ANALYSES OF CANDIDATE CHOICE AMONG
DEMOCRATIC AND REPUBLICAN PRIMARY VOTERS
Variable
Candidate 1 evaluation
Candidate 2 evaluation
Electability
Viability
Chi-Squared
Significance
Percentage of cases
predicted correctly
Democrats
Republicans
(N = 98)
(N = 128)
.791
.637
-.838
-.526
.386
.072
112.7
<.001
.637
-.141
156.2
<.001
92.9
96.1
Note: Entries shown are standardized discriminant function coefficients. For Democrats, candidate 1 = Dukakis, candidate 2 = Gore; for Republicans, candidate 1 = Bush, candidate
2 = Dole. Based on Democrats voting for Dukakis or Gore, Republicans voting for Bush
or Dole.
REFERENCES
Abramowitz, Alan I. 1987. "Candidate Choice before the Convention: The Democrats in 1984."
Political Behavior 9:49-61.
Abramowitz, Alan I., and Stone, Walter J. 1984. Nomination Politics. New York:Praeger.
Aldrich, John H. 1980. Before the Convention. Chicago: University of Chicago Press.
Aldrich, John H., and Cnudde, Charles. 1975. "Probing the Bounds of Conventional Wisdom:
992
Alan I. Abramowitz
A Comparison of Regression, Probit, and Discriminant Analysis." American Journal of Political Science 19:571-608.
Asher, Herbert B. 1976. Causal Modeling. Beverly Hills, CA: Sage Publications.
Bartels, Larry M. 1985. "Expectations and Preferences in Presidential Nominating Campaigns." American Political Science Review 79:804-15.
Bartels, Larry M. 1987. "Candidate Choice and the Dynamics of the Presidential Nominating
Process." American Journal of Political Science 31: 1-30.
Brady, Henry E., and Johnston, Richard. 1987. "What'sthe Primary Message: Horse Race or
Issue Journalism?"In Media and Momentum, eds. Gary R. Orren and Nelson W. Polsby.
Chatham, NJ: Chatham House.
Markus, Gregory B. 1982. "Political Attitudes during an Election Year:A Report on the 1980
NES Panel Study." American Political Science Review 76:538-60.
Markus, Gregory B., and Converse, Philip E. 1979. "A Dynamic Simultaneous Equation Model
of Electoral Choice." American Political Science Review 73:1055-70.
Orren, Gary R., and Polsby, Nelson W. 1987. Media and Momentum. Chatham, NJ: Chatham
House.
Polsby, Nelson W. 1983. Consequences of Party Reform. New York:Oxford University Press.
Alan I. Abramowitz is professor of political science, Emory University, Atlanta, GA 30322.